XPeng launches its robotaxi service in Guangzhou, three years later than the frontrunners and ahead of any other Chinese car manufacturer.

      The first vehicle produced on Monday signifies the beginning of a gradual increase in production. Public pilot programs will commence in the latter half of the year, with fully driverless operations expected by early 2027. On Monday, XPeng announced that its first robotaxi had come off the production line in Guangzhou, positioning itself as the first traditional automaker in China to begin mass production of a robotaxi developed entirely using in-house technology.

      This vehicle is constructed on XPeng’s new GX platform, designed to facilitate Level 4 autonomous driving directly from the factory, rather than requiring aftermarket upgrades. While XPeng's claim is noteworthy, it's important to note that it is not the first company in China to deploy robotaxis; Baidu’s Apollo Go operates over 1,000 vehicles across 22 cities and is the largest commercial operator in the country. Additionally, Pony.ai had a fleet of 1,446 vehicles by late March and aims to reach 3,000 by the end of the year. WeRide also operates more than 1,000 units across over 30 cities and 11 countries. XPeng’s distinction lies in being the first traditional car manufacturer—rather than a solely autonomous-driving firm—to enter series production of a robotaxi in China.

      This distinction combines both marketing and substantive aspects. The pure-play operators have predominantly retrofitted consumer vehicles, often Toyotas or local EVs, with sensor systems and computing resources provided by the AV companies. In contrast, XPeng’s GX-based robotaxi is engineered from the ground up, incorporating the necessary computing power, drive-by-wire chassis, and redundancy for Level 4 operation.

      This robotaxi incorporates four of XPeng’s custom Turing AI chips, yielding a total computational power of 3,000 TOPS, along with a next-generation steer-by-wire system from Bosch that eliminates the traditional mechanical steering shaft. XPeng states that it includes aviation-grade redundancy in safety-critical systems. The software utilized is the company’s VLA 2.0 stack, an integrated vision-language-action model that reduces end-to-end response time to below 80 milliseconds.

      Initial production volumes will be modest. Brian Gu, XPeng's president, informed Reuters last month that the company expects to manufacture hundreds to thousands of robotaxis over the next 12 to 18 months, with pilot operations featuring safety drivers set to begin in the latter half of this year. The goal is to achieve fully driverless commercial service by early 2027. The robotaxi assembly line in Guangzhou possesses a license for intelligent connected vehicle road testing, already conducting Level 4 road tests on public streets.

      The commercial rationale appears partly defensive. XPeng’s push for robotaxis is part of an ambitious diversification plan for 2026 that also includes humanoid robots and a modular flying car, all utilizing the same Turing silicon and vision-language-action software stack. The firm is focused on monetizing its AI capabilities across various formats rather than relying solely on EV profit margins in a declining market influenced by reduced prices and government subsidies.

      Volkswagen represents the most prominent external customer for this strategy. The German automaker acquired a 4.99% stake in XPeng for $700 million in July 2023 and named XPeng’s Turing chip and VLA 2.0 system as this year’s first external commercial customer. This marks the first instance of a major Western carmaker incorporating Chinese-developed autonomous driving software at such a significant level.

      XPeng aims for approximately one million Turing chip shipments in 2026 and intends to open its robotaxi SDK to other fleet operators over time. However, the timing presents challenges. The milestone of starting production aligns with a weaker short-term market forecast. XPeng has projected first-quarter 2026 deliveries to be between 61,000 and 66,000 units, marking a year-over-year decrease of about 30%, which the company attributes to diminished government subsidies and a longer Lunar New Year period.

      The company's Q1 earnings report is due on Wednesday. It recently reported its first quarterly net profit of RMB 380 million for Q4 2025, alongside the downward revision for Q1 forecasts. This development comes at a time when the operational risks associated with large autonomous fleets have become more apparent. In late March, over 100 Baidu Apollo Go robotaxis malfunctioned in traffic in Wuhan due to a software issue, stranding passengers and bringing forth concerns about simultaneous failures in managed fleets—a scenario that traditional vehicle regulations do not adequately address.

      Pony.ai, even after achieving city-level profitability in Guangzhou, has faced similar challenges as it expands. XPeng enters this operational landscape with both a hardware advantage and a proprietary software stack that it has developed and tested. However, it also faces the same governance issues regarding the potential for fleet-scale failures that existing players continue to navigate.

      What XPeng accomplished this week is the initiation of its production line. The first vehicle has been manufactured, and the upcoming pilot operations will